Files
llama3-8b-dpo-4xh100-pilot/all_results.json

9 lines
214 B
JSON
Raw Permalink Normal View History

{
"epoch": 1.0,
"total_flos": 0.0,
"train_loss": 0.6806982485360877,
"train_runtime": 636.997,
"train_samples": 5501,
"train_samples_per_second": 8.636,
"train_steps_per_second": 0.135
}